Why ReiserFS?

[NAS]

Something that I have always been curious about why ReiserFS?

[WeeboTech]

I think in regards to journaling which is faster on Reiserfs and also formatting (which is also faster).

 

With ext3 and journaling support,  data blocks are written twice.

Also very large deletes are much faster with reiserfs.

 

Reiserfs gives me headaches though. I would still prefer to use ext3 so that I can mount disks on my windows machines when needed.

[NAS]

You hit the nail on the head of one of my gripes.

 

My next one is that there are basically no recovery utilities for reiser and many for ext3. OK so its not an exact science but one if the not THE major sellinbg point of unRAID is not striping attracting people like me who want individual drives to still be stand alone operable. Recovery is a natural progression of that.

 

I can see why faster journaling would be an advatage but faster formatting is only a slight advantage considering how comparatively infrequent it happens.

 

Put it another way I suspect if on formatting if unRAID gave an option between reiser and ext3 the majority would stick with the one they all know ext3.

[WeeboTech]

Put it another way I suspect if on formatting if unRAID gave an option between reiser and ext3 the majority would stick with the one they all know ext3.

I would have.

Especially since the author/inventor was arrested and jailed.

[Joe L.]

From what I've read so far, with reiserFS it is fairly easy to extend a partition when upgrading a disk.  Descriptions of how to resize an ext3 partition I've read all seem to start with first turning it into a ext2 partition by removing the journal, and then resizing, and finally restoring the journaling feature.  That added complication might have been one reason for the decision.

 

Tom has indicated he wishes to add support for other file-system types, other than reiserFS.  

I can envision an additional drop-down box per drive on the devices page when initially assigning a drive.  It would allow you to pick from the supported file system types.

Biggest issue is extending the partition size when you upgrade to a larger disk.  (Can NTFS partitions be re-sized on the fly, without affecting the data in them?)

 

Then, need to be able to deal with somebody who decided to change their current flie-system type... should that be permitted, or only a format operation losing all data. (hopefully you will have moved any data to a different drive first)

 

Joe L.

[WeeboTech]

Extending ext3 is easier these days. There have been tools around to do it for a while.

We've been using them for LVM volumes for quite some time. From the last time I did it, You can grow a filesystem live, you can only shrink it while unmounted.

[bubbaQ]

Especially since the author/inventor was arrested and jailed.

 

How is this relevant?  It is a FOSS project, and has many hands in the pie.  I don't give a rat's pajamas if Hitler and and the KKK invented it or named it... let's stick to technical merit.

[jimwhite]

(Can NTFS partitions be re-sized on the fly, without affecting the data in them?)

Yes... the Disk Management module allows this....

 

 

[WeeboTech]

Especially since the author/inventor was arrested and jailed.

 

How is this relevant?  It is a FOSS project, and has many hands in the pie.  I don't give a rat's pajamas if Hitler and and the KKK invented it or named it... let's stick to technical merit.

 

I would care if the main developer/driving force of a project was locked away. People in his company were concerned.

It is relevant enough for some major distributions to alter their default OS from reiserfs back to ext3.

Last I read reiserfs was just in maintenance mode.  Ext4 is on the horizon. 

[Billped]

From ZDNet:

 

"Prior to the whirlwind and highly publicized trial, Hans Reiser and his small team were working on Reiser4, but much of this development ground to a halt due to his legal woes, and the project is more than likely to die an unfortunate death by virtue of its lead programmer having to serve a minimum 25 year life sentence in prison."

 

http://blogs.zdnet.com/BTL/?p=8647

 

 

Bill

[limetech]

At the beginning we chose ReiserFS for these reasons:

 

1. No need to think about i-node allocation & no possibility of running out of i-nodes & then trying to explain to a user why there might be free space left on a disk but not able to create any more files on it.

 

2. Very easy expansion of the file system when the partition expands, which is what happens when a user replaces a smaller disk with a larger one.

 

3. Since journalled, no lengthy 'fsck' upon unclean shutdown.

 

4. Exceptional small file performance, and very good large file allocation - early on I extensively tested this and found that for larger files such as video, reiserfs tended to keep the blocks almost always contiguous, or at least large sections were contiguous.

 

5. Comparatively fast 'mkfs' and fast 'mount'.

 

6. Existence of Windows tools to read off the data if necessary.

 

Other file systems were considered:

ext2 - potential problems running out of inodes, slow fsck, no journal.

ext3 - also potential problems running out of inodes, not quite mature enough for me (at the time, circa 2005-2006).

xfs - no windows tools (still none I think)

ntfs - no stable driver at the time

 

If only Hans and his wife could have "just got along" I'm confident Reiser4 would probably be the pre-eminent FS in the linux world today.  But because of the stigma, companies are distancing themselves, which I guess I can't blame them.

[NAS]

Tom if you had to choose again today in light of all the development and "news" what would you choose?

[bubbaQ]

If you want to suggest a change because of a likelihood of no future development, then that is legitimate.... but whether the loss of the lead programmer is due to death, prison, or a decision to join a cloistered monastery, is irrelevant.

 

There are people who want to continue development.... I'll wait to pass (technical) judgment after seeing what they produce (or don't produce).  There is likely to be a name change, however.  To allude to my previous post, I'd use the Hitler File System or the KuKluxKlan file system if it was good ... but I'd also have no problem if they changed the name since the name is not a technical feature.

[NAS]

..I'd use the Hitler File System or the KuKluxKlan file system if it was good ...

 

I get the spirit of what your saying but those examples are truly horrible and have no place here.

 

I would reiterate that IMO lack of seasoned recovery tools for reiser make it one of my least favourite options. I also wonder if the inode issues of ext3 would really effect real world NAS users.

 

 

[hypyke]

No love for ZFS around here?

[NLS]

..I'd use the Hitler File System or the KuKluxKlan file system if it was good ...

 

I get the spirit of what your saying but those examples are truly horrible and have no place here.

 

Why I heard HitlerFS has nice recovery tools and manages to reuse space quite effectively.

 

Note that since I am Greek and my country suffered both Italian and ON TOP of that, 3rd Reich occupation and whole villages got wiped in WWII, I should be sensitive too.

But I wouldn't worry with such examples. Humanity has (hopefully) matured quite a bit to be able to make fun with its own mistakes of the past. (interestingly enough - my job's closest business partners are German and Jewish!)

 

I would use a HitlesFS or a KKKFS if it was any good, but I would probably call it "that filesystem" to forget about the name and creators and be more politically correct. Ah... and would try to find a pirated version without any guilt.

 

---

 

Indeed the question remains. What would Tom choose today?

I wouldn't bother with a switch unless we loose something really vital some other more modern FS gives and would really make unRAID better.

 

A good place to note here (again) that before finding out about unRAID, I too had a quite relative idea with Tom, only it would be "low level" FS agnostic (and even mix and hide under the cover of the top layer FS). As long as an FS driver is there... ...Sure would make things more difficult with far more variables to take care... but things could slowly move towards that target and maybe unRAID 2010 should be that way.

 

 

[WeeboTech]

Please use the accurate and correct name in posts. Lets not confuse, antagonize or push anyone's buttons.

 

As far as ZFS, If it were in the kernel I would consider it.

 

At this point, My wish would be to allow us to use reiserfs or ext3.

I have found reiser to be faster. It's just that the windows too I want to use is insufficient for my needs.

Furthermore all my other linux boxen  use ext3.

[NLS]

"It's just that the windows too I want to use is insufficient for my needs."

 

Sorry didn't get this phrase.

 

 

[NAS]

..

Furthermore all my other linux boxen  use ext3.

 

Thats a valid point. I would hazard a guess that most users have all their other system on either NTFS or ext3 which should add to the + side of this equation.

 

From a commercial point of view and my assumptions on the userbase i would suggest that the negative PR that reiserfs has gotten is - side of this equation.

 

I can honestly say that if i was coming to this fresh i would probably have walked away at the point i read reiserfs because of the press and an opinion that is based on no fact that reiserfs project in its current carnation is dead. In essence i probably wouldnt have read any further on the website.

 

 

[Joe L.]

"It's just that the windows too I want to use is insufficient for my needs."

 

Sorry didn't get this phrase.

 

 

Probably just a typo

"It's just that the windows tool I want to use is insufficient for my needs."  (in other words, the reiserfs driver is read-only on windows)

[PhilH]

I just want the file system that is the most stable & works with unRAID the best. If that is ReiserFS then so be it. I don't really care if the developer is in jail.

[SSD]

I second that!  BTW I have had good luck with reiserfsck recovering data.

[WeeboTech]

I would like a choice that is all.

 

The windows tool for resierfs is an explorer like tool.

There is an ext2 driver for windows that lets you use it like a regular file system. It's saved my butt a few times and also allows me to move very large vmware images between work and home and/or use them directly.

 

I did mention that the reiserfs does have many speed advantages.

As far as the author's legal state... It's relevant to the point of future development.  Time will tell.

 

[SSD]

I don't disagree with a providing choices.  Its just a question of priority.  Since reiserfs has performance and ease of implementation advantages, and it is already working, I am not sure providing that choice now when there are so many feature requests that would benefit everyone are on the list.

 

It does seem likely that reiserfs should eventually be replaced.  Certainly seems the author got what he deserved - but my moral compass does not require immediate cessation of using the FS.  If he had attempted genocide, I'd likely feel differently.  I would be supportive of adopting a new standard as reiserfs3 features become dated.  I'd suggest something other than reiserfs4, even is reiserfs4 came out and had some minor advantages, given the author's status.

 

This is all my opinion.

[bubbaQ]

Several of the key developers of Reiser FS have suggested a name change for the project .... some others are resisting on principle -- you don't invalidate good work someone did because in some other arena, they did something bad.  Sports records are not taken away because of crimes unrelated to the sport.  Had Hitler been the first man to scale Everest, he would still be the first man to scale Everest.  Had Einstein been a murderer, we wouldn't stop using his work.

 

Tom articulated VERY sound reasons for the choice, and pointed out significant hurdles to overcome in a change.  That's the end of it for me.  There are adequate tools for RFS, both for recovery and mounting under other OSes.  Quantity of utilities doesn't mean quality nor does a lesser quantity mean inadequacy.  The suggestion for expanding the supported file systems in unRAID has been made, and Tom has acknowledged it as a long-term goal, and explained sound technical reasons for the choices made.  Let's leave it at that.